neuron (wip)
intro
in my final year of university, i selected as an elective a course on Computational Neuroscience, overseen by Dr Simon Stringer. at that time, my primary focus had been passing exams, on ‘just getting through’ the uni ordeal at large, and i’d never truly found myself engaged with much of the compulsory content (psycholinguistics in particular). the only moments i’d sincerely engaged were when a lecturer, with some unorthodox style, would out themselves as their own species of raging genius, only present in front of us to complete the otherwise burdensome pedagogical aspect of his or her professorship. these were either the best, or most dull lectures of the week. fortunately, Dr Stringer’s were of the former type.
his course centred on a single axiom: that any computational system, if seeking to replicate a process already mastered by the brain, must borrow (principles, structure, plasticity) from the brain’s own architecture to achieve it. Dr Stringer must’ve repeated statements of this kind just about once a minute. but, as we delved deeper into the architecture of the visual system, with its seemingly geological layers stacked one on top of another, and confronted the complexities of visual object detection within convolutional neural networks, i realised at some vague point that i was investing my full attention in this science – and doing so with more focus than at any other time in my academic career. so it was, that in studying the brain with Dr Stringer, i was able to realise my own brain’s latent intellectual potential and begin to engage with the world at large with greater attention and sensibility. this seemed to unlock something – and it’s no coincidence that around this time, or not long after it, my pre-teen interests in literature, in learning languages, and in making any ‘art’ whatsoever, emerged out of their stasis, and thrust themselves back into the regular fold of my life.
thanks to the joys of studying Dr Stringer’s hyperfixations, and despite his (quiet) pivot to developing defence tech, i’ve since kept an inconsistent, but nonetheless well-engaged eye on the neuroscience field, whenever i’ve had the time. only now, three years since i was last in a classroom, however, have i come to the decision that i should start writing down whatever insights i find/enjoy. so, i present here the beginnings of a very basic and completely amateur effort at introducing the field of study known as ‘Neuroscience’.
drawings by Santiago Ramón y Cajal (public domain), via Wikimedia Commons
functional simulation (& artist’s rendition) of a leaky integrate-and-fire neuron
modelled in python, uses τ·du/dt = −(u − u_rest), goofy ‘firing’ animation is for effect
each click is an input from another neuron, and nudges the voltage up. but the membrane leaks, so every nudge fades away within a second or so. click slowly and the voltage never gets anywhere; click fast enough and the inputs pile up past threshold, and the neuron fires.
right after a spike it can't fire again: clicks during the spike do nothing (grey ×). the voltage then dips below rest before recovering, so the next spike takes more work. timings are slowed right down so it's playable. under the hood this is a leaky integrate-and-fire neuron: between clicks, τ·du/dt = −(u − u_rest). the spike's shape is drawn by hand, because the model itself only knows that a spike happened.